Implicit conversion from float -21.01549053356281 to int loses precision